Spectral extrapolation for ultra-wide band radio frequency super-resolution tumor localization in the breast.

Abstract

The use of ultra-wide band (UWB) radio frequency (RF) as an alternative to x-ray mammography for the detection and localization of breast tumors has been an area of focused research over the last decade. Unlike x-rays, UWB RF is non-ionizing and poses no risk of inducing cancer in examined patients. However, the reduction in operating frequency of UWB RF compared to x-rays results in much poorer localization when using classical space-time adaptive signal processing. This work investigates the synergistic use of a temporal signal spectrum extrapolation technique and contemporary basis pursuit de-noising (BPDN) beamforming to achieve medically relevant tumor localization accuracy within a heterogeneous simulated breast volume. If the beamforming process is viewed as a matched filtering operation, the spectrum extrapolation technique artificially increases the bandwidth of the received signal to afford much sharper correlations and hence contributing to increased localization accuracy. BPDN also contributes to increased localization accuracy by employing L1-norm constraints which shows marked clutter suppression effects in this context.